Very bad/thirsty i7 930 or am I just not doing it right?

Bluesman · 11 Apr 2010 at 11:27

I have been doing my head in for the past couple of weeks since I put my new x58/i7 system together.

Basically I just can't seem to get 4GHz 'stable', and I have already reached my voltage limit... I really am quite shocked at the amount of volts I have had to put into this chip... It's batch no. is 3951A683 and I actually had high hopes for it since a batch very similar to that (last digit only difference) is an extremely good OCer and figured mine would be similar. So don't know if I am just an incredibly unlucky person or my inexperience with i7 OCing is having an effect.

Anyway so 4GHz was the first thing I went for right off the bat. Using a popular i7 OC guide I tweaked the following settings: IOH voltage - 1.20, ICH voltage - 1.20, CPU PLL - 1.88, QPI voltage - 1.35, vcore - 1.275v (in bios).

These settings caused a bsod (124 code) before even one pass of LinX. From research I learnt that the 124 error code indicates QPI voltage being too high/too low. So I trying a couple of voltage variations with no luck. The only thing that did the job.... was more vcore... a lot more... It took 1.320v (cpu-z) to get 25 passes of LinX. Thing is.. prime would not last more than 45 minutes (odd blank screen freeze)... Which I found strange given that I can play games for hours on end with no issues at all.. not to mention the 25 passes of LinX.

3.8GHz/1.28v (could possibly go lower) on the other hand passes both (prime 12hrs).

What am I doing wrong? And is there anything that can be done to allow me to lower the vcore and give me more OC headroom in the process?

Here are my current BIOS settings for 4GHz (which will not pass prime):

Jumperfree settings

CPU ratio - 21.0
Speedstep - Disabled
BCLK - 191
PCIE Freq - 100
DRAM Freq - 1531MHz
UCLK Freq - 3036MHz
QPI Link Data Rate - Auto

CPU voltage - 1.33125
CPU PLL Voltage - 1.88
QPI/DRAM voltage - 1.30000
IOH Voltage - 1.20
IOH PCIE Voltage - 1.50
ICH Voltage - 1.20
ICH PCIE Voltage - 1.50
DRAM Bus Voltage - 1.64

LLC - Enabled
CPU Diff. Amplitude - 800mV
CPU Clock Skew - Auto
CPU Spread Spectrum - Disabled
IOH Clock Skew - Auto
PCIE Spread Spectrum - Disabled

CPU Config

C1E Support - Disabled
Hardware Prefetcher - Enabled
Adjacent Cache Line Prefetch - Enabled
Intel Virtualisation Tech - Enabled
CPU TM - Enabled
Execute Disabled Bit - Enabled
Intel HT - Enabled
A20M - Disabled
Speedstep - Disabled
C-STATE Tech - Disabled

Any help/advice would be greatly appreciated.

Nodiggidy123 · 11 Apr 2010 at 11:35

knock the ram up to 1.66v. althoug h yu are already running the ram lower than spec so i doubt it. but it could be worth a try. also tinker a bit with the qpi/dram voltage. sometimes less volts can make it stable.

Bluesman · 11 Apr 2010 at 11:42

Nodiggidy123 said:
knock the ram up to 1.66v. althoug h yu are already running the ram lower than spec so i doubt it. but it could be worth a try. also tinker a bit with the qpi/dram voltage. sometimes less volts can make it stable.

I think I may have tried that already but I'll do it again anyway just in case.

Also should I be setting the 'QPI Link Data Rate' to anything in particular? Couldn't find much info on it.

EDIT: Yeah I was just going to ask about messing with the QPI/DRAM voltage some more. I don't think I have tried it at stock settings yet. Thanks.

DIABLO · 11 Apr 2010 at 12:18

Your settings look spot on as you said the instability was cured by upping the vcore so as you said you have a thirsty chip. You should get away with lowering the QPI voltage though as usually people don't need 1.35v to get 191MHz.

Bluesman · 11 Apr 2010 at 13:48

Well looks like I actually managed to make some progress as far as getting prime stable.

Here's what I did.

DRAM voltage from 1.64 - 1.66
QPI Link Data Rate - from 'Auto' to 64_ _ something, can't remember. Lowest setting anyway.

Before I was getting a freeze during prime at pretty much 45 mins in every time. Now, with the above settings tweaked, it's been running for about 1hr 20 mins so far.

If this prime run passes 6+ hrs, I guess I can try lowering the vcore, or perhaps even raising the clocks? Was QPI Link Data the key to my OC issues?

EDIT: Forgot to mention, I also discovered that my auto ram timings were not correct in cpu-z. They read 8-9-9-24, when they should have been 9-9-9-24. So I also set that right too.

Justintime · 11 Apr 2010 at 14:06

I know its probably an overreaction on my part, but i'd keep the RAM voltage down to under 1.65v

Bluesman · 11 Apr 2010 at 14:59

Justintime said:
I know its probably an overreaction on my part, but i'd keep the RAM voltage down to under 1.65v

Seems the opinon is very mixed on this to me.

Is there not some form of vdroop in memory voltage though, making it safe to set to 1.66? Also if it's not likely the increase to 1.66v actually solved my prime stability problem (and it was the QPI change) then I will put it back to 1.64. What are the thoughts on this?

Btw, I just stopped the prime run at 2hrs 30mins. No errors, no issues. I plan to do a more lenghtly run later on (overnight).

DIABLO · 11 Apr 2010 at 15:56

My DDR3 automatically sets the voltage to v1.65 on my RAM when it's run at 1600MHz and the voltage is left on auto. Yours is probably the same.

thingemajib · 11 Apr 2010 at 17:10

never actually found out why >1.65v ram will frazzle an i7/i5...

twist3d0n3 · 11 Apr 2010 at 19:04

thingemajib said:
never actually found out why >1.65v ram will frazzle an i7/i5...

i'm sure it's to do with the memory controller being on chip, and so being of 45nm process, so the qpi voltage and dram voltage directly effect that part of the processor

DIABLO · 11 Apr 2010 at 21:20

Never heard that one guys.... All i know is that the reccomended volts for the QPI is 1.35v, 1.65v for RAM and that you must keep the difference between the two at 0.5v or lower.

Sayso · 11 Apr 2010 at 21:27

take everything off auto...when i clocked on a rampage left the RAM voltage on auto and it was using 1.75

edit: it sounds like you are just putting voltages you have seen somewhere and hoping they work...clock the damn thing properly spend some time on it and you will get it stable.

alex24 · 11 Apr 2010 at 21:36

Every chip is different. I would say don't jump straight to 4GHz, work slowly up, making sure each clockspeed is FULLY stable before continuing. It will take a lot more time, but it will be a much better method to make sure you're optimising the settings as best as possible, and learning about the nuances of your chip.

VortX · 11 Apr 2010 at 21:50

Triffic, I've got the same batch as you haha, let us know how you get on!

Bluesman · 11 Apr 2010 at 21:59

Yeah I suppose you're right guys. I should have taken things slower and worked up. Guess I was just too hungry for some instant horse power!

Anyway so as I explained in my last post, things seem to be looking up as far as getting my current OC prime stable. Let me reiterate - The 4GHz OC was LinX stable from the beginning, the main issue was prime.

Will be starting another prime run tonight to run through till the morning, and basically hoping for the best. If I get a full 8-12hrs prime run with no errors, then I can deem my OC LinX and prime stable. Then I can start trying to reduce some voltages and hopefully (best case scenario) be able to push my OC further.

The setting which I believe had the positive effect was QPI Link Data. Perhaps it's previous 'Auto' setting was messing things up.

VortX, will do!

Nodiggidy123 · 11 Apr 2010 at 22:05

Bluesman said:
The setting which I believe had the positive effect was QPI Link Data. Perhaps it's previous 'Auto' setting was messing things up.

yes this! afaik the qpi link should be fixed at double the speed of your ram. i had stability issues untill i realised this.

Sayso · 11 Apr 2010 at 22:15

clocking an I7 ( i haven't owned a 930 but have had 4 920's) isnt hard but getting it "right" with the best temps/voltages can take some time.
for instance my rampage was stable at 4.4 within 10mins of getting it put together at 1.35V and other random numbers...but i finally settled at 4.2 @ 1.28 and a 15C temp drop after tweaking!!

Bluesman · 11 Apr 2010 at 22:45

Nodiggidy123 said:
yes this! afaik the qpi link should be fixed at double the speed of your ram. i had stability issues untill i realised this.

Yeah I wasn't aware of that at all. I thought it was just the UCLK Frequency that needed to be 2x the ram freq. I had heard nothing about the QPI Data Rate while looking at guides. I think different terms were being used to refer to the same setting, and that's how it slipped over my head.

alex24 · 11 Apr 2010 at 22:53

Indeed, prime (blend in particular) is sometimes a tough one to get stable, even when other things are working fine.

Jokester · 12 Apr 2010 at 00:06

Set PLL to 1.8V and check stability again.